You are on page 1of 38

The Volume and Structure of the Transatlantic Slave Trade: A Reassessment

Author(s): David Eltis


Source: The William and Mary Quarterly , Jan., 2001, Vol. 58, No. 1, New Perspectives
on the Transatlantic Slave Trade (Jan., 2001), pp. 17-46
Published by: Omohundro Institute of Early American History and Culture

Stable URL: https://www.jstor.org/stable/2674417

REFERENCES
Linked references are available on JSTOR for this article:
https://www.jstor.org/stable/2674417?seq=1&cid=pdf-
reference#references_tab_contents
You may need to log in to JSTOR to access the linked references.

JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide
range of content in a trusted digital archive. We use information technology and tools to increase productivity and
facilitate new forms of scholarship. For more information about JSTOR, please contact support@jstor.org.

Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at
https://about.jstor.org/terms

Omohundro Institute of Early American History and Culture is collaborating with JSTOR to
digitize, preserve and extend access to The William and Mary Quarterly

This content downloaded from


139.47.113.28 on Fri, 07 Jan 2022 09:38:16 UTC
All use subject to https://about.jstor.org/terms
r~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~r

4~~~.3.( 94 ~ *"wo 9

I~~~~~~~~~~~~~~~~~~~~~~~~~~I

I ft.
Ih
e a a a a a a a a a a a a~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~

Map I: The Atlantic World circa 3750. Based on 'The Atlantic


1998). vi-vii. by Ira Berlin. Reprinted by permission of the
Harvard College.

This content downloaded from


139.47.113.28 on Fri, 07 Jan 2022 09:38:16 UTC
All use subject to https://about.jstor.org/terms
* iveroi -

)L don EUROPE
FRANCE

*Nantes

*Bordeaux

AFRICA

Cape Coast Castle X


%I/nCristianborg

Axim4~'~ Bonny
OdCalahar
o erado Po
Princip6

Sa oe. >t Loango


*Cabinda
~.Mpinda......

*Luanda

irst Two Centuries of Slavery in America (Cambridge, Mass.,


Press. Copyright ? 1998 by the President and Fellows of

This content downloaded from


139.47.113.28 on Fri, 07 Jan 2022 09:38:16 UTC
All use subject to https://about.jstor.org/terms
I ~~ ~; )a I art)r V

0 ('E Nzrv>

.I1..f\ I1 K.. I- I- - ll

K,'~ o 'i _ '.44


I~~~~~~~~~~~~~~. I _ I <

Ma I:Wet fic n heEa f thernalni

and ports of embarkation, based on David Eltis, Stephen D. Behrendt, Da


eds., The Trans-Atlantic Slave Trade: A Database on CD-ROM (Cambr
Wrenn.

This content downloaded from


139.47.113.28 on Fri, 07 Jan 2022 09:38:16 UTC
All use subject to https://about.jstor.org/terms
BAHAMAS
A TL

/' % \ '' OCE

~~~~~~~UA Usd

[ \~SPANISH LSE
PUERTO
RICO

J ^ J~~~~~~AM1AICA SAINT DOMsINGUE

\CENTRAL Caribbean ANTII.LE

M~~~t 0 )

Km. 0 ) s W 1,w \

Map III: The Catibbean


Trans.Atdank Siasw Tre

This content downloaded from


139.47.113.28 on Fri, 07 Jan 2022 09:38:16 UTC
All use subject to https://about.jstor.org/terms
W+E
N

A TLANTIC

OCEAN

PUERTO
RICO

1MINGUE

LESSER

ANTILLES ,

'URAQ 7 BARBADOS

of thetansatanticslavetrad s on ltIstAD
of the transatlanti slave trade, basedon li ta. d

This content downloaded from


139.47.113.28 on Fri, 07 Jan 2022 09:38:16 UTC
All use subject to https://about.jstor.org/terms
)rmp \ t ) %%III WI '1 1Imtg S
~~~~~~~1 ii HHIISI S $t:
S I t tell\ 1%- I kilts I I 'A H 1 )
0
MSiliticmi

IRI \i II

IN[) \ RI) TLK I V 7TI(1


Carib b C?11 Sc a Ilk 51 r Sliqic)(

?2 (rcnt . 1.,.,~,

DI i0t I C K1141i %N @

c9 _______'t" " -"

Map IV: The Lesser Antilles in the Eighteenth Century, based on Eltis et al.,
eds., Trans-Atlantic Slave Trade Database. Map drawn by Rebecca Wrenn.

This content downloaded from


139.47.113.28 on Fri, 07 Jan 2022 09:38:16 UTC
All use subject to https://about.jstor.org/terms
S
C ~~\ l'(urt~rl '+1~l

: ~~~~J A M A I C A

Km. /lot)(d

Map V: Jamaica circa 1770, sh


of the transatlantic slave trade
Database. Map drawn by Rebec

This content downloaded from


139.47.113.28 on Fri, 07 Jan 2022 09:38:16 UTC
All use subject to https://about.jstor.org/terms
The Volume and Structure of the
Transatlantic Slave Trade: A Reassessment

David Eltis

S INCE work on The Trans-Atlantic Slave Trade: A Database on CD-ROM


began in the late i980s, questions about it have fallen overwhelmingly
into two categories: does the dataset have names of African individuals
or groups, and by how much will it change estimates of the number of peo-
ple forced into the traffic. The answer to the first question-no-is easier to
give than the answer to the second. Paradoxically, the new data will probably
modify currently accepted estimates of the size of the trade less than it will
change knowledge of most other aspects of the trade.1 Links between Africa
and the Americas, deaths of both slaves and crew on the voyage, the age and
sex of slaves, national participation in the trade, almost any organizational
question, ownership patterns, and many other topics will draw on the new
collection to a much greater degree than will the long debate over how many
Africans arrived in the Americas between 1519, the likely date of the first, and
i867, the probable year of the last transatlantic slave voyage direct from
Africa.2 Nevertheless, the data do support a revised aggregate estimate, and,
more important, they provide the basis for more accurate assessments of who
carried the slaves, from which part of the African coast they embarked, and
where in the Americas they were taken. What follows is the first report of a
full-length independent reassessment of the size and distribution of the traf-
fic currently in preparation.
This reassessment of the volume and structure of the transatlantic slave
trade is a culmination and extension of the work of others. Philip Curtin and
Joseph Inikori built on a combination of estimates of people who lived

David Eltis is a professor of history at Queen's University, Research Associate at the


W.E.B. Du Bois Institute of Harvard University, and Research Lecturer at the University of
Hull. He thanks the National Endowment for the Humanities, the Social Sciences and
Humanities Research Council of Canada, the Andrew Mellon Foundation, the W.E.B. Du Bois
Institute, and Hull University for financial support. He thanks Stephen D. Behrendt, Pieter
Emmer, Stanley L. Engerman, David Richardson, and Lorena S. Walsh for comments on earlier
versions of this article and Sarah Hughes for help in checking references.
I David Eltis, Stephen D. Behrendt, David Richardson, and Herbert S. Klein, eds., The
Trans-Atlantic Slave Trade: A Database on CD-ROM (Cambridge, i999) (hereafter cited as Eltis
et al., eds., Trans-Atlantic Slave Trade Database).
2 Before 1519, all African slaves carried into the Atlantic disembarked at Old World ports,
mainly Europe and the offshore Atlantic islands. From 1493, the year of Columbus's second voy-
age, some of these slaves or their progeny entered the New World. The first vessel carrying
slaves to have sailed directly between Africa and the Americas appears to have arrived in Puerto
Rico in 1519.

William and Mary Quarterly, 3d Series, Volume LVIII, Number i, January 200i

This content downloaded from


139.47.113.28 on Fri, 07 Jan 2022 09:38:16 UTC
All use subject to https://about.jstor.org/terms
I8 WILLIAM AND MARY QUARTERLY

through the slave trade era, aggregated shipping data, and population pro-
jections of recipient regions in the Americas, and Paul Lovejoy consolidated
the conclusions of scholars who had recovered more new data from the
archives after Curtin published his findings; the present reassessment, by
contrast, is based on individual voyages for the busiest parts of the trade,
nearly half of which have only recently become known.3 For the most part,
this work uses voyage-by-voyage shipping data rather than the estimates of
then-contemporaries and historians derived from those data. As information
becomes available and as assumptions underlying the estimates are refined,
the conclusions offered here will change. Moreover, the new estimates have
yet to be tested systematically against the known demographic data, a
process that also leads to some revision.4
The CD-ROM does not, and its successors never will, record every voy-
age that set out to obtain slaves. That it contains a majority of those voyages
is certain, for three reasons. First, internal checks are possible for some
branches of the trade, which suggests fairly complete coverage. Thus French
vessels, when they returned to their home port after delivering slaves in the
Americas, typically included references to other French vessels met with on
the voyage. More than 95 percent of French ships cited in this way were
already in the dataset.5 Second, in the last stages of preparation of the British
data, after incorporation of all other major sources, Lloyd's Lists, perhaps the
most comprehensive and independent single source for eighteenth-century
British ship movements, was combed. For the years for which copies of
Lloyd's Lists survive, data on 8,ooo British slaving voyages exist. This ship-
ping gazette provided a rich haul of additional material on voyages already
known, but added only 35o new voyages to the set. Reassuringly, more of
this increment sailed from London, for which sources are weaker, than from
other British ports.6 Gaps exist in all sources, but for both the French and
British slave trades there are no years or ports for which no coverage exists.7

3 Philip D. Curtin, The Atlantic Slave Trade: A Census (Madison, i969); Joseph E. Inikori,
"The Known, the Unknown, the Knowable, and the Unknowable: Evidence and the Evaluation
of Evidence in the Measurement of the Trans-Atlantic Slave Trade" (paper presented at the
Williamsburg conference on the Trans-Atlantic Slave Trade Database, Sept. 1998), s-3; Paul E.
Lovejoy, "The Volume of the Atlantic Slave Trade: A Synthesis," Journal of African History, 33
(1982), 473-501, and "The Impact of the Atlantic Slave Trade on Africa: A Review of the
Literature," ibid., 30 (1989) 365-94; Per 0. Hernaes, Slaves, Danes, and the African Coast S
The Danish Slave Trade from West Africa and Afro-Danish Relations on the Eighteenth-Century
Gold Coast (Trondheim, Nor., I995) 140-71.
4 To encourage debate and make it transparent, the spreadsheets underlying this article are
posted on the Omohundro Institute/ William and Mary Quarterly website <http://www.wm.
edu/oieahc> where readers can view the derivation of the estimates and indeed try new assump-
tions (and data) of their own. See also the 7 tables at the website.
5 Jean Mettas, Repertoire des Expeditions Negrieres Franfaises au XVIIIe Siecle, 2 vols. (Paris,
1978-1984).

6 For a full explanation see the Introduction to Eltis et al., eds., Trans-Atlantic Slave Trade
Database.
7 Thus, Inspector General Thomas Irving's comments that the naval offices in the various
British West Indian islands failed to send in all quarterly returns (cited in Inikori, "The Known,
the Unknown, the Unknowable," 12-13) is not significant given the numerous alternative
sources for the British trade between i672 and i807. Sixteen separate sources are listed for some

This content downloaded from


139.47.113.28 on Fri, 07 Jan 2022 09:38:16 UTC
All use subject to https://about.jstor.org/terms
VOLUME AND STRUCTURE OF THE SLAVE TRADE I9

A third check on completeness builds on confidence in the French and


British data. Generally, the regions in which French and British vessels
disembarked their slaves are known, as are the export crops that those
slaves and their descendants produced, these being the main activities
supported by the slave trade. It is possible therefore to compare planta-
tion output from such regions with that of regions where less is known
about their share of the trade. If output is substantially lower in the
region where the slave trade is less recorded, then it is unlikely that the
slave trade to such areas was greater than the slave trade to the French
and British Americas. Such comparisons at least provide an order of
magnitude for slave arrivals.8

This article takes four cumulative steps in drawing the broad out-
lines of the trade. First, it generates estimates of national participation.
Second, it distributes these national groupings across African regions on
the basis of large samples of known regions of slave purchases. These
distributions are then summed to derive quarter-century totals for each
African region. Third, it performs a similar exercise for regions of dis-
embarkation in the Americas. Fourth, it outlines the African origins of
slaves disembarking in a few American regions. Space constraints dictate
that only a few of the American regions can be treated in this way, pro-
viding merely a taste of what the dataset is able to do in constructing
transatlantic links.9
Table I provides estimates of national participation in the trade as
expressed in the numbers of slaves carried from Africa under each
national flag. The national affiliations of 23,302 (86 percent) voyages are
known. For a further 2,465, the context of the voyage and the name of

British voyages, and even in the s7th century, sources per voyage reached double figures
and derive from Africa and Europe as well as the Americas.
8 To convert shipping data into estimates of slaves carried, two adjustments are
required. For a minority of voyages, no information is available on the outcome. In such
cases, the vessel left its home port intending to obtain slaves but cannot be traced further.
In other cases, the vessel is known to have reached Africa and even obtained slaves, but no
further information is available. Such vessels, composing about io% of voyages in the
dataset, are assumed to have delivered slaves to the Americas, and this decision almost cer-
tainly imparts some upward bias to estimates of the volume of the trade presented here. A
second adjustment is necessary because only 72% of the vessels that took slaves on board
left a record of the number of slaves either embarked or disembarked. For such vessels an
imputed number of slaves is added by averaging numbers that are known by regions of
embarkation and disembarkation. Full information on these procedures is provided in the
SPSS programs contained in Trans-Atlantic Slave Trade Database, though these have been
modified slightly since publication.
9 Note that the numbers in this article do not always accord with those in the dataset
on account of additions and corrections made since the dataset appeared. Brazilian-
registered vessels, owned overwhelmingly by Portuguese citizens, are grouped with those
sailing under the Portuguese flag. Generally, it is not always easy to tell the difference
between English colonial vessels and English vessels and, later, between U.S. and British
vessels where no explicit identification is available. All such pre-i808 voyages of unknown
flag are taken to have been British colonial if sailing before 1776 and U. S. if sailing after
1775 (effectively after 1782, given the impact of war on North America shippers).

This content downloaded from


139.47.113.28 on Fri, 07 Jan 2022 09:38:16 UTC
All use subject to https://about.jstor.org/terms
20 WILLIAM AND MARY QUARTERLY

the ship, its owner, or its captain make possible inferences about place of
registration. Some vessels, responding to nineteenth-century attempts to
suppress the traffic, especially after 1844, would sail without registration
papers or would switch flags during a slaving voyage. Other vessels, regis-
tered in one country but belonging to nationals of another, would carry out
the whole voyage under false papers, although overall this is not a large
group (probably less than I percent). These are assigned to the Portuguese
flag.10 After all adjustments, the British (including British colonials) and the
Portuguese account for seven of ten transatlantic slaving voyages and carried
nearly three quarters of all slaves who embarked in Africa. Broadly, the
Portuguese dominated before i640 and after 1807, with the British displacing
them in the intervening period. The French, Dutch, Spanish, Danish, and
United States slave traders appear as bit players by comparison, though at
particular times some of these national groups did assume importance. Each
national trade is taken up briefly in turn.
The British trade is, after the French, the most thoroughly researched of
all national branches of the traffic. For the years before 1713, after 1779, and
for Bristol, Whitehaven, and all minor ports in Lancashire and Cheshire
throughout the slave trade era, published estimates already exist that are
broadly consistent with the dataset."I For the years 1714 to 1779, a new count
for London, Liverpool and ports in the West Country, and Scotland is made
directly from the set. For London and the British outports, the proportion of
slaving voyages included in the set is at least 90 percent.'2 Almost no slavers
leaving Liverpool in the eighteenth century escaped notice, as indicated by
the extraordinary richness of the sources for the port, which include planta-
tion shipping registers, Admiralty passes, Colonial Treasurers' accounts, port
books, and newspapers.'3 Overall, the dataset probably contains some record

10 Thus, examining only the registration or flag at the outset of the voyage may yield a
misleading picture of national participation in the i9th-centuLry trade.
II Behrendt, "The Annual Volume and Regional Distribution of the British Slave Trade,
I780-1807," J. African Hist., 38 (1997), 187-211; Richardson, Bristol, Africa and the Eighteen
Century Slave Trade to America, vol. I: The Years of Expansion, r698-1729 (Gloucester, i986); vol.
z: The Years of Ascendancy, 173o0-1745 (GlouLcester, 1987); vol. 3: The Years of Decline, 1746-1769
(Gloucester, 1991); vol. 4: The Final Years, I770-I807 (Gloucester, 1997); Richardson and
Maurice M. Schofield, "Whitehaven and the Eighteenth Century British Slave Trade,"
Transactions of the Cumberland and Westmorland Antiquarian and Archaeological Society, 92
(1992), 183-204; Schofield, "The Slave Trade from Lancashire and Cheshire Ports Outside
Liverpool, c. 1750-c. I790," in Roger Anstey and P.E.H. Hair, eds., Liverpool, the African Slave
Trade, and Abolition: Essays to Illustrate Current Knowledge and Research, enlarged edition
(Liverpool, i989), 239-81; Eltis, "The British Transatlantic Slave Trade Before 1714: Annual
Estimates of Volume and Direction," in Robert L. Paquette and Engerman, eds., The Lesser
Antilles in the Age of European Expansion (Gainesville, Fla., i996), and "The Volume and African
Origins of the Seventeenth-Century English Trans-Atlantic Slave Trade: A Comparative
Assessment," Cahiers dEtudesdAfricaines, 138 (1995), 617-27.
12 Some 234,000 slaves were carried on vessels coming from unknown British ports. Most
of these sailed from Bristol, London, or Liverpool. Although the dataset includes almost all
ships leaving these ports, the particular port of departure is not always identified in the set.
13 Richardson, Katherine Beedham, and Schofield, Liveipool Shipping and Trade, 1744-1786
(ESRC Archives, University of Essex, i992). See the sources in Eltis et al., eds., Trans-Atlantic
Slave Trade Database, for other references.

This content downloaded from


139.47.113.28 on Fri, 07 Jan 2022 09:38:16 UTC
All use subject to https://about.jstor.org/terms
VOLUME AND STRUCTURE OF THE SLAVE TRADE 21

of 95 percent of the voyages that sailed from Britain between 1714 and 1779.
To allow for missing records, slave departures on Liverpool ships between
1714 and 1779 are divided by o.99 (thus assuming a i percent omission) and
those on London and outport vessels by o.9 (thus assuming a io percent
omission). Integration of the new and previously published estimates yields
the series for the British trade in column i of Table 1.14
Records for some other national groupings of slave traders are as or even
more complete. For the French trade, dividing the number of slaves moved
on French ships in the dataset by 0.95, at least after 1707, will compensate
for missing data. The results appear in column 2 of Table I. Before 1707,
records are much less complete, but the French slave trade also operated far
below its eighteenth-century peak in these years.15 The post-i675 Dutch traf-
fic has been catalogued and analyzed by Johannes Postma, who believes that
some record of almost every Dutch slave vessel has survived.'6 Lloyd' Lists
contains records of only six Dutch slave voyages not in his set, and most of
these vessels were captured or shipwrecked before completing their voyages.
Before i675, it is possible to make a partial reconstruction of Dutch slaving
activity, which got underway with the Dutch occupation of northeastern
Brazil in the i630s.17 Column 3 of Table I shows the results. The small
Danish trade is well documented. Per Hernaes estimates 97,850 slaves leavin
Africa but includes 8,700 leaving Christiansborg on non-Danish ships,
almost all of which are likely counted under the flags of other nations. The

14 For the detailed derivation of this series, see Eltis, Behrendt, and Richardson, "The
Transatlantic Slave Trade: A New Census" (Cambridge, forthcoming). Some British-owned ves-
sels sailed from French ports tinder the French flag prior to the outbreak of war in 1793 without
first leaving a trace in the British records. Similar activity is likely in the very small Dutch and
the even smaller Spanish trades late in the i8th century. The records of the host country nor-
mally list such vessels as domestically owned. Allowances for such vessels would increase slightly
the overall estimate of all these trades. The estimates in Table I already allow for missing vessels,
and no further adjustment appears called for.
15 A few French slave vessels not in Mettas-Daget show up in Lloyd's Lists. In most cases
such ships were captured or wrecked before reaching the Americas. In other words, their move-
ments were less likely to be recorded in French sources than vessels that completed their voy-
ages. For the French slave trade during the Peace of Amiens see Eric Saugera, "Pour une histoire
de la traite francaise sous le Consulat et l'Empire," Revue Franfaise d'histoire d'Outre-Mer, 76
(1989), 203-29. For discussion of the adjustments to the pre-1707 data see Eltis, Behrendt, and
Richardson, "Transatlantic Slave Trade: A New Census."
16 Johannes Menne Postma, The Dutch in the Atlantic Slave Trade, i600-i815 (Cambridge,
1990), 110, ii8, 304-411.
17 For the unpublished work of Franz Binder, on which these estimates of the early
are based see Postma, Dutch in the Atlantic Slave Trade, 31-35. From i634 to i648, the Dutch
occupied Pernambuco, the largest Brazilian sugar-growing area, and from i630 to i654 they held
other Brazilian ports. From i630 to i65i, Dutch ships carried 26,i26 slaves into what had been
Pernambuco and surrounding territory; Ernst Van den Boogaart and Pieter Emmer, "The Dutch
Participation in the Atlantic Slave Trade, 1596-i650," in Henry Gemery and Jan S. Hogendorn,
eds., The Uncommon Market: Essays in the Economic Histoly of the Atlantic Slave Trade (New
York, 1979), 369. This series replaces the one in H. Watjen, Das Hollandische Kolonialreich in
Brasilien: ein Kapital aus der Kolonial Geschicte des I7. Jahrhunderts (Berlin, 1921), 487, widely
cited in the older literature. For further adjustments to these estimates see Eltis, Behrendt, and
Richardson, "Transatlantic Slave Trade: A New Census."

This content downloaded from


139.47.113.28 on Fri, 07 Jan 2022 09:38:16 UTC
All use subject to https://about.jstor.org/terms
22 WILLIAM AND MARY QUARTERLY

remaining 89,15o Africans probably accoun


ing voyages made in the period i66o-i8o6.18 The other northern powers
played a much smaller role. Brandenburg and then the Hanse towns con-
tributed a few voyages, Swedes a few more, perhaps amounting to five thou-
sand departures overall.19 Total estimates are 94,200 departures (78,600
arrivals) for the northern powers.20 The quarter century breakdown of these
estimates is shown in column 6 of Table I.
Less information is available for other slave trades. From 1714 to i8ii,
the dataset contains records of 1,500 voyages setting out, or probably setting
out, from North America.21 Some of these were European-based voyages
that returned directly to the African coast after delivering their first group of
slaves to the Americas. The majority are from Rhode Island ports.22 A com-
bination of Lloyd's Lists, colonial treasurers' accounts and naval office lists of
various British colonies, and a series of journals of ship movements kept by
officials of the British Gold Coast fort of Cape Coast Castle (all far removed
from New England) provides an independent check on the completeness of
the North American mainland trade.23 These sources suggest that the Rhode
Island trade amounted to about i,000 voyages, that Rhode Island ports sent
out about half of all slave vessels from the North American mainland,
though vessels setting out from most ports outside New England were larger
and carried more slaves than New England ships, and finally that between
1730 and 1807 one transatlantic slaving voyage was based in the English
Caribbean for every four on the North American mainland. Thus about
2,000 slaving voyages left North American mainland ports between 1714 and
1807 carrying an estimated 220,600 slaves (and disembarking 205,500).
Before 1714, North American slave vessels are combined with the larger
British trade. Given an estimated 2,000 mainland voyages, 500 Caribbean

18 Svend Holsoe and Hernaes, working independently of each other, have collected what
they consider to be close to comprehensive data on this branch of the traffic. Encouragingly,
their estimates are similar. See Hernaes, Slaves, Danes, and the African Coast Society, 170-233;
Holsoe to author, Copenhagen, January i998. Voyage mortality appears to have been relatively
high on Danish vessels at i6.4% (N = 86, Standard Deviation [SD] = 13.9), and this implies
arrivals numbering 74,600. Eltis et al., eds., Trans-Atlantic Slave Trade Database, by contrast,
has details of 6i,397 slaves leaving Africa and 51,28i arriving in the Americas. The distribution of
the Hernaes/Holsoe aggregates by quarter century is based on the distribution in these data.
19 Adam Jones, Brandenburg Sources for West African History, 1680-1700 (Stuttgart, 1985),
i-II. There are records of i6 ventures under these flags, mainly during the era when the electors
of Brandenburg controlled Gross-Friedrichsburg just west of Cape Three Points on the Gold
Coast between i683 and 1721; the fort was virtually moribund for the first two decades of the
i8th century before the Dutch took it over.
20 Mortality on non-Danish vessels sailing before i8oo is taken at 20%.
21 Information on some vessels assumed to have left North American ports is imputed
from the name of the captain or data on previous voyages.
22 Most of these appear in Jay Coughtry, The Notorious Triangle: Rhode Island and the
African Slave Trade, 1700-1807 (Philadelphia, 1981), 241-85, drawn overwhelmingly from Rhode
Island sources such as newspapers and local official records.
23 The T70 series, vols. 1463, 1472, 1476, 1484, 1498, 1515, 1517-19, 1522-32, 1534-38,
1544-59, 1561-74, in the British Public Record Office, contains the Cape Coast Castle reports of
vessel movements, and the C028 series contains the colonial treasurers' accounts.

This content downloaded from


139.47.113.28 on Fri, 07 Jan 2022 09:38:16 UTC
All use subject to https://about.jstor.org/terms
VOLUME AND STRUCTURE OF THE SLAVE TRADE 23

slaving voyages are estimated to have sailed after 1730, and an additional 50
for the years 1714 to 1730, when the dataset actually contains more slave ships
sailing from the Caribbean than from mainland colonies. Except for
European-based vessels returning directly to Africa for more slaves, these ves-
sels were generally quite small. Overall, 550 Caribbean-based voyages are
estimated for 1714-1807, carrying 59,400 slaves from Africa (and disembark-
ing 51,300).24 Thus, vessels from ports in the British Americas carried off
280,000 slaves after 1713 (see Table I, column 5). For the non-English and
non-Portuguese Americas, transatlantic slave ship departures were very few,
and the historians of the French, Dutch, and Danish slave trades have likely
caught all such voyages in their archival nets.
The Portuguese and Spanish trades, unlike the British and Dutch,
focused on a narrow range of areas in the Americas. The winds and currents
of the South Atlantic ensured that the slave trade to Brazil was the preserve
of traders based in the major Brazilian ports mainly Portuguese, even after
Brazilian independence. Spanish slavers traded only to the Spanish Americas,
mainly Cuba, and accounted for a small proportion of pre-i8io transatlantic
arrivals.25 Very few Spanish vessels appear to have been active before 1790,
and the dataset shows i6,ooo slaves in total leaving in Spanish vessels, all to
Cuba. To allow for missing records, this figure is increased by half, and total
departures of 24,000 (20,800 arrivals) are thus projected for 1700-i8io.26
The Spanish traffic, mostly Cuban-based, expanded rapidly from these mod-
est levels after i807 (see Table I, column 4).
Although the Portuguese trade is the most underrepresented in the
dataset, it nevertheless constitutes a quarter of it, and if the estimates of the
Luso traffic below are correct, then more than half of all Portuguese slaving
voyages are included. All transatlantic vessels, except for the very few identi-
fied under other flags, are assumed to have been Portuguese until i640. For
the very earliest period before i6oo Curtin's estimate of 50,000 arrivals

24 Applying the same loss ratios as those estimated for Rhode Island ships means that 550
voyages obtained slaves in Africa, and 533 of these voyages reached the Americas. The derivation
of the estimates of total slaves carried is in Eltis, Behrendt, and Richardson, "Transatlantic Slave
Trade: A New Census."
25 From the first two centuries of the trade, the Spanish chose to issue licenses or asientos
that allowed mostly non-Spaniards to bring slaves to their territory rather than to carry slaves
themselves. The licenses often changed hands several times before coming into the possession of
the actual shippers of slaves. In the early days the asientistas could be Spanish, Genoese, or
Portuguese, but by the late i6th century were normally Portuguese. As it is difficult to separate
out these different groups, all non-Spanish transatlantic shippers to the Spanish Americas before
I700 are labeled Portuguese here. Given the small scale of the early trade, the upward bias this
imparts to the Portuguese trade is minor. In the second half of the i8th century, occasional
Spanish slaving voyages occurred. Under the I778 Treaty of Pardo, Spain acquired the island of
Annob6n in the Bight of Biafra from Portugal for the express purpose of establishing a slaving
base (see the memo dated Feb. 26, I840, F084/383, fol. 262, PRO, and the documents in
F084/299, fols. I9-25, PRO, apparently removed from the Spanish archives, for a summary of
this activity). As these efforts were not successful, a substantial transatlantic Spanish traffic did
not develop until the British and Americans pulled out of the trade in i807.
26 A small sample of shipboard mortality on Spanish vessels before i8ii suggests I3.7% of
those taken board failed to reach the Americas (N = ii, SD = i6.6).

This content downloaded from


139.47.113.28 on Fri, 07 Jan 2022 09:38:16 UTC
All use subject to https://about.jstor.org/terms
24 WILLIAM AND MARY QUARTERLY

in Brazil still holds. For the Spanish Americas, a major adjustment to his
estimate of 75,000 arrivals for the same period is required. Bullion exports to
Spain expanded rapidly at the end of the sixteenth century and peaked
shortly after i6oo. The best recent estimate of slave arrivals in Spanish
America between 1595 and i640 is 268,200, and no fewer than 30 percent of
these arrived in the six years, 1595-i600.27 This end-of-century bulge, which
brought 80,5oo Africans to Spanish America easily the largest annual aver-
age to any major national jurisdiction in the Americas before the explosive
growth of the British sugar sector is the basis of doubling Curtin's estimate
for the Spanish Americas from 75 to 150 thousand. This implies a total of
200,000 arrivals in the Americas in the sixteenth century. Average voyage
mortality for 1595-i640 for sixty-six Portuguese vessels was 25.6 percent (20
percent estimated for the shorter route to Brazil) and supports an estimate of
departures from Africa of 264,i00 between 15I1 and i6oo.
For the seventeenth century, the trade to Spanish America switched
from Portuguese to Dutch and English control in midcentury, and thereafter
the Portuguese slave trade is virtually synonymous with the traffic to Brazil,
except for a short renewal of the asiento between i696 and 1701, this time on
the part of the Portuguese Cacheu Company.28 From i6oi to i640, arrivals
on Portuguese vessels in the Spanish Americas totaled i87,700 (the survivors
of 252,000 departures). For Brazil, older estimates posited 200,000 arrivals
in Brazil between i6oo and i65o or 173,700 after deducting those coming
on Dutch ships and 325,000 between i65o and 1700.29 Analogies with bet-
ter documented branches of the slave traffic suggest that both earlier esti-
mates are too high. The major buyers of African slaves in Brazil before 1700
were sugar planters. The expansion of their sugar sector in the second half of
the sixteenth century was probably similar to that of its later English coun-
terpart. Both sectors changed their supply of labor to predominantly African

27 All data in the set, as well as the estimate are from Enriqueta Vila Vilar, Hispanoamerica y
el comercio de esclavos (Seville, I977). Eltis et al., eds., Trans-Atlantic Slave Trade Database, has
83,200 arrivals-29,800 in I595-i600.
28 Under the new contract, perhaps 5,000 slaves entered Spanish America. See Georges
Scelle, "The Slave-Trade in the Spanish Colonies of America: The Asiento," American Journal of
International Law, 4 (igio), 6i2-6i. For the difficulty of obtaining slaves from sources other
than the British and the Dutch after i66o see Leslie B. Rout Jr., The African Experience in
Spanish America, I502 to the Present Day (Cambridge, I976), 44-48.
29 Frederic Mauro's 35-year-old review of the evidence is still basic, though the shipping
and demographic roots of his discussion are not robust. He estimated io8,ooo arrivals in
Pernambuco between i6oo and i652-75,000 before i630-all iln Portuguese vessels. His esti-
mate post-i630 is roughly consistent with the Dutch data. Allowing for Bahia and Rio de
Janeiro-the only other significant ports of entry-he estimated 200,000 slave arrivals from
i6oo to i65o. After i65o, Mauro fell back on Goulart's aggregate estimate of 20,000 slaves pe
year arriving in the Americas, of whom he thought 6,000-7,000 could have disembarked inl
Brazil. This implies an estimate of 325,000 arrivals between i65o and I700. See Mauro,
Portugal et L'Atlantique au XVIIe Sikcle (Paris, i960), I74-8I, and Mauricio Goulart, Escrivaddo
Africana no Brasil (das origens a extincdo do trafico) (Sdo Paulo, I950), ii6. Stuart Schwartz, Sugar
Plantations in the Formation of Brazilian Society (Cambridge, I985), 342, accepts Goulart's figure
of 2,000 per annum for Bahia alone, which is broadly consistent with Mauro's discussion.

This content downloaded from


139.47.113.28 on Fri, 07 Jan 2022 09:38:16 UTC
All use subject to https://about.jstor.org/terms
VOLUME AND STRUCTURE OF THE SLAVE TRADE 25

slaves, the first from indigenous and the second from a European indentured
servant regime. The growth of the English plantation system was likely more
explosive and its focus on African labor at the end of the transition greater.
At some point around i675, exports of sugar from the English Caribbean sur-
passed the volume of Brazilian sugar exports just prior to the i624 Dutch
attack. One hundred thousand slave arrivals in Brazil between i6oo and i625
(half the 200,000 currently projected for i6oi-i650) compares well with the
better documented 140,000 who arrived in the British Caribbean between
i640 and i675, even after allowing for the Indian labor that Brazilian planters
used. After i625, stagnation in the Bahian sugar sector suggests that i00,000
slave arrivals in the second quarter of the seventeenth century are unlikely,
even including arrivals on Dutch vessels. Pernambuco had accounted for well
over half of Brazilian sugar output before occupation by the Dutch. Total
arrivals in Portuguese vessels in the second quarter of the seventeenth cen-
tury should probably be halved to 50,000 Thus, 150,000 is allowed for the
Portuguese traffic to Brazil, i6oo-i650 (making i87,500 departures, on the
basis of an estimate of 20 percent mortality on the shorter South Atlantic
route). The addition of 252,000 departures for the Spanish Americas and
i87,500 for Brazil yields 439,500, the total number of slaves carried to the
Americas in Portuguese vessels, i6oi-i650.
For i650 to 1700, the estimate of 325,000 arrivals in Brazil derived from
Frederic Mauro and Mauricio Goulart also seems in need of downward revi-
sion.30 The English sugar sector absorbed 177,000 enslaved Africans, yet
Barbados and the English Leewards together likely yielded more plantation
produce than the whole of Brazil in 1700.31 At the end of one hundred years
of slave arrivals and a plantation economy that grew slowly between i65o and
1700, why would Brazil have taken in more slaves than the English
Caribbean? It is possible that slaves were used more extensively in nonexport
sectors in Brazil than in the Caribbean, but African slaves were purchased,

30 More recent estimates imply an even larger volume-nearly half a million slaves enter-
ing Brazil on Portuguese ships in this half century. Patrick Manning (using Pierre Verger's data
on tobacco ships for West Africa) has suggested 2,300 slaves a year from the Slave Coast to
Brazil; "The Slave Trade in the Bight of Benin, i640-I890," in Gemery and Hogendorn, eds.,
Uncommon Market, I07-4I. Joseph C. Miller estimates total slave departures from West Central
Africa at ii,8oo per year in "The Number, Origins, and Destinations of Slaves in the Eighteenth-
Century Angolan Slave Trade," in Inikori and Engerman, eds., The Atlantic Slave Trade: Effects on
Economies, Societies, and Peoples in Africa, the Americas, and Europe (Durham, N. C., i992), i09, and
in Way of Death: Merchant Capitalism and the Angolan Slave Trade i730-i830 (Madison, i988),
233. Miller's graph yields approximately iio,ooo for the i66os, I30,000 for the i670s, I25,000 for
the i68os, and I05,000 for the i69os. Deducting known departures in Dutch and English ships
of about 2,300 a year from Angola (Postma, Dutch in the Atlantic Slave Trade, Iio, II2-I3; and
Eltis, "African Origins and Value of the English Slave Trade," table 4) from these West Central
African figures leaves an annual traffic to Brazil of more than ii,000 (Manning, 2,300, and
Miller, 9000+) a year or c. i0,000 arrivals per year for 50 years.
31 Eltis, "British Atlantic Slave Trade," i96-98. The Leewards received slaves from Dutch-
held St. Eustatius that are not included here. More than offsetting this inflow were sales from
Barbados to the Spanish Americas and French Windwards. On sugar output, Barbados and the
English Leewards exported produce worth an estimated ?684,800 in I700. Bahia exported an
estimated ?227,400 and Brazil as a whole an estimated ?568,5oo See Eltis, The Rise of Afric
Slavery in the Americas (New York, 2000), i98-99.

This content downloaded from


139.47.113.28 on Fri, 07 Jan 2022 09:38:16 UTC
All use subject to https://about.jstor.org/terms
26 WILLIAM AND MARY QUARTERLY

one way or another, with Brazilian exports, and export trends provide some
guide to the potential for obtaining slaves. An estimate for Brazil of 177,000
is accepted for this half century. An allowance of 12.5 percent voyage mortal-
ity yields 202,300 departures from Africa.32 Including the illegal Portuguese
traffic to Spanish America and the activities of the Cacheu Company
(i0,000 arrivals), then the total Portuguese trade for i651-1700 becomes
i87,000 arrivals in the Americas, implying 214,800 departures. Because the
Brazilian sugar sector began to expand once more in the final quarter of the
century, when the Bahia-Bight of Benin slave route began, three quarters of
these departures (i6iIoo) are assigned to i676-1700 and the rest to
i65i-i675.
For the long eighteenth century, the Portuguese slave trade is once more
defined by the slave trade to Brazil, and fortunately, the sources are stronger
here than for the previous century. For Bahia from i678, the CD-ROM
incorporates vessels departing for the African coast with licenses to carry
tobacco (in effect a slave ship departure series) and includes other data
besides. As a result of Luso-Dutch treaties in i64i and i66i, tobacco vessels
were required to leave io percent of their cargo at Elmina before they pro-
ceeded to the Slave Coast, and Dutch records of these tobacco deposits fill
in some of the years for which Bahia tobacco licenses have not survived.33
Indeed, just twelve years between i678 and i8io have no information on
slave ship movements from the port of Bahia. Nine of these years fall
between 1717 and 1725, and the probable reason for the gaps is the loss of
historical records, not the absence of slave trading. Interpolation fills the
gaps.34 Aggregate departures total 656,ooo between 170i and i8io, and, after
allowing for shipboard mortality, arrivals are 598,200.35 For the four quar-

32 Shipboard mortality on vessels sailing the shorter passage between the Slave Coast and
Bahia was below that for the West African-Caribbean route. A proposal to establish a Portuguese
company at Corisco Island, c. I724, intended to displace or reduce Portuguese slave trading on the
Slave Coast stated Io% loss as "etant la computation ordinaire" (quoted in Stowe ms., ST 28, 7,
Huntington Library, San Marino, Calif.). When data for the Bahian trade become available in the
early i9th century, mortality loss was about half that on other routes. Mortality on ships arriving
in Bahia from the Bight of Benin, mainly from I8Io-I814, was 4.6% (N = ioi, SD = 7-5).
33 Verger published a count of Bahian vessels granted licenses to carry tobacco to the
African coast, which is also in effect a count of slave ships, in "Mouvement des navires entre
Bahia et le Golfe du B6nin (XVIIe-XIXe siecle)," Revue Francaise d'Histoire d'Outre-Mer, 55
(i968), io-i2, and Trade Relations between the Bight of Benin and Bahia, I7th to ipth Century,
trans. Evelyn Crawford (Ibadan, I976), 24-26. See Eltis et al., eds., Trans-Atlantic Slave Trade
Database, for additional sources, including Biblioteca Nacional do Rio de Janeiro, Documentos
Histdricos da Biblioteca Nacional do Rio de Janeiro, IIO vols. (Rio de Janeiro, I929-I955), esp.
vols. 6i, 62. These contain data for I7I5 and the first 3 months of I7i6 recorded shortly before
the originals were lost or destroyed.
34 To convert what are for the most part records of possible ship departures into a series of
slaves leaving Africa, two steps are required. The first is to compute a mean for slaves carried on
vessels leaving the Slave Coast with known numbers on board, readily calculated from Eltis et
al., eds., Trans-Atlantic Slave Trade Database. The second is to make allowance for the years for
which data are missing. For 3 of the i2 missing years, quinquennial averages are assigned, and
for the 9-year block of missing years, the figures are estimated on the basis of a simple linear
interpolation between 5-year averages on either side of the gap.
35 Voyage mortality data for the pre-i8ii years are few, and the rate from Angola to Rio de
Janeiro between I70i and i8io is used instead (8.8i%, N = 326, SD = 7.2).

This content downloaded from


139.47.113.28 on Fri, 07 Jan 2022 09:38:16 UTC
All use subject to https://about.jstor.org/terms
VOLUME AND STRUCTURE OF THE SLAVE TRADE 27

ters of the eighteenth century, departures for Bahia are estimated at 2i8,300,
114,700, 103,500, and 123,400.
Portuguese traffic to the rest of Brazil drew mainly on Angola. Official
records of ship departures as well as annual summaries of slave departures
exist for most years for the Angolan ports Luanda and Benguela.36 For
Luanda, data are missing for just four years 1715-1717 and 1732 and it is
possible to add interpolations for these as well as to extrapolate backward for
the eight years before 1709 when the summary series begins. Such procedures
yield total departures for 170i-i8i0 of 959,300, or 144,100, 236,500, 217,300,
and 237,200 for the four quarter centuries 170i-i800, required for Table I.
Total arrivals are estimated at 874,800.37 For Benguela, the other major
Portuguese port in Angola, annual totals exist for all but ten years between
1730 and i8io.38 All years that lack data fall in the decade and a half after the
annual series begins, and it is possible that lack of data here simply reflects
lack of slave trading. There is no evidence of slave trading activity before
1730, and until 1745, slave departures averaged just i,000 annually. Some
activity likely did occur, and 5,ooo departures from Benguela are assigned
each of the quinquennia 1726-1730 and 1731-1735. Thus departures from
Benguela are taken as continuous from 1726. They aggregate to 301,900
(275,300 estimated arrivals), I70I-i800.39 Benguelan departures by quarter
century are zero in the first quarter, then 26,600, 99,300, and 176,i00.
How many slaves were smuggled out of these ports and thereby eluded
the officials on whom the above data rely? The dataset offers some check on
smuggling. Between 1796 and i8io it is possible to match two sets of annual
estimates of slaves leaving Luanda and Benguela on Portuguese ships, one
made on the African side of the Atlantic, the other on the Brazilian side.40
Rio de Janeiro records report 8I,200 slaves leaving the two principal African
ports, whereas Angolan records report the departure of 88,700. Given that
Pernambuco and Bahia also drew slaves from Angola at this time about
one-fifth of the vessels arriving in Luanda between 1796 and i8io originated

36 For trading routes see Miller, "Numbers, Destinations, and Origins of Slaves in the
Eighteenth-Century Angolan Slave Trade," 78-89, and Goulart, A Escraviddo, 203-05. For a
review of these summaries, see Jose C. Curto, "A Quantitative Reassessment of the Legal
Portuguese Slave Trade from Luanda, Angola," African Economic History, 20 (i992), I-25, and
"The Legal Portuguese Slave Trade from Benguela: A Quantitative Re-Appraisal," Africa, i6
(I993), ioi-i6.
37 Mean voyage mortality (as a % of those embarked) was 8.8i% between Angola and Rio
between I70i and i8io (N = 326, SD = 7.2).
38 Sources are Klein, "Portuguese Slave Trade," 9i8, for I738-I74I, I744, I747-I800, and
Curto, "Slave Trade from Benguela," for I730, I742, and i8oi-i8io. See also Miller, "Legal
Portuguese Slave Trading from Angola: Some Preliminary Indications of Volume and
Direction," Rev. FranC. d'Hist. d'Outre-Mer, 62 (I975), I35-76, and "Number, Ori
Destinations of Slaves," for I780-I783 and I785-I8Io. For departures from Benguela, this last
article offers a different series from the one printed in the I975 article. I use the later source here.
39 The Rio-Angola voyage mortality rate of 8.8i% is used.
40 The Angolan data are reported by Miller in "Number, Origins, and Destinations of
Slaves," 92, ioo-oi. The Rio data in the set are discussed in Klein, The Middle Passage:
Comparative Studies in the Atlantic Slave Trade (Princeton, I978), 5I-94.

This content downloaded from


139.47.113.28 on Fri, 07 Jan 2022 09:38:16 UTC
All use subject to https://about.jstor.org/terms
28 WILLIAM AND MARY QUARTERLY

in Pernambuco and Bahia these totals are close.41 Large-scale smuggling in


this later period, if it existed at all, must have circumvented officials on both
sides of the Atlantic.42 A relatively modest io percent supplement is allowed
to cover illegal slaves on Portuguese ships (and illegal Portuguese ships), and
the estimates of quarter century departures from Luanda and Benguela given
above are adjusted accordingly. How many slaves left on Portuguese vessels
from ports in Angola other than Luanda and Benguela? French, British, and
Dutch vessels were all active in Angola from the i670s until i807. The
dataset indicates that the northen Angola coast, or points on and north of
the Congo River, was dominated by northwestern European nations. Of 356
slave vessels arriving at Rio between 1795 and i8io, twenty-one reported
from Africa north of Luanda, though all except one of these voyages
occurred after i807 in other words, after the British, Dutch, and French
had moved out of the trade.43 Portuguese departures from north of Luanda
were therefore rare enough to be ignored.
Finally, some assessment is required for three minor branches of the
eighteenth-century Portuguese slave trade. The traffic in small vessels from
Cacheu and Bissau to Amazonia is estimated at 57,ooo departures (53,000
arrivals), 1,500 in each of the first two quarters of the century, 14,300 in the
third, 24,800 in the fourth, and 14,800 in i8oi-i8ho.44 Elsewhere, Portuguese
ships began to obtain slaves in southeast Africa from the mid-179os. The Rio
de Janeiro records indicate 14 ships carrying 5,300 slaves from Mozambique
between 1796 and i809 (1,378 before i8oi) and 4,0oo arrivals.45 There, too,
the Portuguese moved in only after the St. Domingue slaves and the revolu-
tionary wars had diverted the attention of the French. In addition, the
Companhia Geral de Pernambuco e Parafba carried 9,500 slaves from the
Slave Coast mainly to Pernambuco between 1760 and 1782 (8,6oo estimated
arrivals).46 Little is known about the extent of noncompany slave traffic to

41 See graph 3 in Miller in "Number, Origins, and Destinations of Slaves," I50. The shares
of Brazilian ships arriving in Luanda, I796-i8io, from Rio de Janeiro, Bahia, and Pernambuco
(the only 3 ports included in the graph) do not total one, and I have assumed that Benguela was
inadvertently omitted.
42 Smuggling is never costless. Ports exist because they offer facilities for importing and
exporting goods. Avoiding such facilities is always expensive. If established ports are used for
smuggling, then officials must be bribed. See the discussion in John J. McCusker, Rum and the
American Revolution: The Rum Trade and the Balance of Payments of the Thirteen Continental
Colonies, 2 vols. (New York, I989), I:302-I7, that concludes that scholars generally exaggerate
the problem of smuggling in i8th century Atlantic ports.
43 Lisbon vessels attempted to slave there in the I790S but obtained few slaves and had to
complete their complement in Luanda; Miller, Way of Death, 623. See also the discussion in his
"Number, Origins, and Destinations of Slaves," 92.
44 For the derivation of these totals see Eltis, Behrendt, and Richardson,"Transatlantic
Slave Trade: A New Census." They draw on Colin M. MacLachlan, "African Slave Trade and
Economic Development in Amazonia, I700-I800," in Robert B. Toplin, ed., Slavery and Race
Relations in Latin America (Westport, Conn., I974), 112-45, and Mettas, "La Traite Portugaise
en Haute-Guin6e, I758-I797,' J. African Hist., i6 (I975), 343-63.
45 For the Rio de Janeiro end of this activity see Klein, Middle Passage, 56, and Manolo
Florentino, Em Costas Negras: Uma Historia do Trafico Atldntico de Escravos Entre Africa e o Rio
de Janeiro (Sdo Paulo, I997), 78-84, 234. For the Mozambique records of this activity see
Edward A. Alpers, Ivory and Slaves in East-Central Africa (London, I975), I85-93.
46 Ant6nio Carreira, As Companhias Pombalinas, 2d ed. (Lisbon, I983), 247-49.

This content downloaded from


139.47.113.28 on Fri, 07 Jan 2022 09:38:16 UTC
All use subject to https://about.jstor.org/terms
VOLUME AND STRUCTURE OF THE SLAVE TRADE 29

Pernambuco from 1778 to i8io. An estimate of 30,000 departures (27,000


arrivals) is allowed for this minor branch of the traffic. The total of these
disparate trades is 101,700 departures (92,700 arrivals), 1,500, 1,500, 21,259,
and 48,176 for each quarter century after 1700 and 29,263 for i8oi-i8io. The
Portuguese trade can now be summarized. The totals for each of the four
quarters from 1701 to i8oo are the sums of, first, departures for Bahia on ves-
sels that brought out tobacco; second, departures for Brazil from Luanda and
Benguela (increased by IO percent to allow for smuggling); and, third, the
three minor branches of the trade just discussed, with 328,800 for the
i8oi-i8io subsection (see Table I, column 7).47
The discussion so far has focused on the traffic before i8ii. Estimates of
the size of the slave trade after i8io have been made elsewhere, and the prob-
lem is to distribute these estimates by nationality.48 Ships often flew flags or
carried registration papers that were bogus, and sometimes they carried sev-
eral sets at once. A breakdown of known nationalities of slave vessels in the
post-i8io data suggests that ii percent of slave ships in the set were French,
i8 percent were Spanish, and 71 percent Portuguese. Total departures are
estimated at ii,062,000 and arrivals at 9,599,000.49
How does Table I's grand total of ii million departures (9.6 million
arrivals in the Americas summed from Table III) compare to existing esti-
mates? Lovejoy's last survey of the secondary work on this topic in 1989
yielded ii,863,000 Africans entering the Atlantic slave trade and "9.6-io.8
million slaves would have been imported into the Americas." His "pre-
ferred" total of arrivals is 1o.2 million. Inikori suggests that recent archival
research and a different configuration of Lovejoy's sources yields i2,689,000
departures from Africa. Per Hernaes's recent, independently computed esti-
mate is just IOO,OOO higher than Inikori's.50 These compare with Curtin's
1969 estimate of 9,566,Ioo arriving in the Americas and the Old World wit
a 20 percent margin of error. Curtin's estimate of arrivals in the Americas

47 Spreadsheet and supporting tables are at OIEAHC website.


48 Carreira, As Companhias Pombalinas, 250-54.
49 Departures from Africa after i8io are estimated at 2,738,900 and arrivals in the Americas
at 2,383,800. Eltis, Economic Growth and the Ending of the Transatlantic Slave Trade (New York,
I987), 250-54, has 2,289,400 arrivals. The difference between total departures from Africa and
total arrivals in the Americas is not just voyage mortality but slaves diverted to European
colonies in Africa by the various naval squadrons. Recaptured slaves are included in the new
arrival figures. After I840, many slave ships began their voyage in U. S. ports, flew the U. S. flag
until they embarked slaves, at least, and were prosecuted successfully in U. S. courts, but none
of this activity establishes U. S. ownership. Ownership of these vessels and the human cargoes
themselves remain for the moment unidentified. U. S.-registered slave ships are here distributed
proportionately between Spanish and Portuguese flags on the ground that most were in reality
owned by citizens of Spain and Portugal.
50 Lovejoy, "Volume of the Atlantic Slave Trade"; Inikori, "The Known, the Unknown,
the Knowable, and the Unknowable," 2-3; Hernaes, Slaves, Danes, and African Coast Society,
I40-7I, offers new global estimates of arrivals of io,869,594, to which he adds 55,889 libertos to
Sdo Tome and Principe for I876 to i900. He then converts to exports by adding IS% mortality
and gets I2,787,758.

This content downloaded from


139.47.113.28 on Fri, 07 Jan 2022 09:38:16 UTC
All use subject to https://about.jstor.org/terms
30 WILLIAM AND MARY QUARTERLY

implies ii.2 million departures from Africa.51 These aggregations, however,


include the traffic to Europe, the Mascarene Islands, and the offshore
Atlantic islands, all of which Table I excludes. Coerced migrants to these
excluded areas amounted to about 200,000, and adding this to the Table I
total means that the new total is slightly above Curtin's figure. More
broadly, it may be taken to accord with his aggregate figures, though, as dis-
cussed below, the distributions over time and space differ from his.
In addition, the new figures are more certain than earlier estimates.
Specifically, Curtin's confidence interval might be reduced from + 20 per-
cent to perhaps half that range. Given the search of sources on the slave
trade during the past five years it is unlikely that another million or more
slaves could have been removed from Africa beyond those allowed for here-
even in the less-researched Portuguese trade. Work on the more obscure
branches and periods of the Portuguese slave trade is not likely to result in
large increases in the present estimates, although it will certainly fill in gaps.
Recent detailed search in the British sources has lowered the estimates of the
British slave trade. Future work on the Portuguese trade may well show
allowances for the unknown to have been as overly generous as they are now
known to have been in the British trade. Even if this is not the case, the CD-
ROM should make it possible to focus the debate over volume and direction
on vessel-by-vessel shipping data rather than comparisons of opinions of his-
torical figures who either had access to that data or who observed the conse-
quences of slave trading.52
Future reassessments are not likely to threaten the picture of overall
Portuguese dominance in the trade apparent in Table I. Yet that dominance
is to some extent misleading. During the era of serious British participation,
i647 to i807, British slavers carried five slaves across the Atlantic for every
four in Portuguese vessels. If we include colonial vessels, the margin of
British superiority was closer to one-third than one-fourth. Given the
Portuguese shift into the Cuban trade after the British outlawed slave trad-
ing, as well as the wide-open British territories in Guiana and Trinidad, it is
hard to believe that the British dominance over the Portuguese in every
quarter from i651 to i8oo would have faded after i807 without British aboli-
tion. In every branch of the trade where the British competed directly with
the Portuguese, the latter lost out. Portuguese dominance came about only
because the British were late getting into the trade and then were among the
first nations to abolish it. The major difference between the two leading car-
riers is that the British trade was based largely in England, whereas the
Portuguese was anchored in Brazil. At the peak of the British traffic in the
1790s, a large slave vessel left England every other day. Moreover, large ports
came to dominate the traffic as it reached its acme. The organizational bases

51 The overall estimate is actually for arrivals in the Americas and the Old World. For
these estimates see Curtin, Atlantic Slave Trade, and Lovejoy, "Volume of the Atlantic Slave
Trade," 473-50I, and "Impact of the Atlantic Slave Trade," 365-94, quotation on 373.
52 Thus, for example, differences between Inikori and me on the number of voyages in the
early i8th-century English slave trade should be resolved easily on a voyage-by-voyage basis.

This content downloaded from


139.47.113.28 on Fri, 07 Jan 2022 09:38:16 UTC
All use subject to https://about.jstor.org/terms
VOLUME AND STRUCTURE OF THE SLAVE TRADE 31

of the Portuguese trade were spread around the South Atlantic basin, with
Pernambuco, Bahia, and Rio de Janeiro more important than Lisbon. In the
second half of each of the i82os, i830s, and i840s, more than one hundred
slave ships a year left Rio, and the scale of activity was not far off what it had
been in Liverpool fifty years earlier.
The size of the traffic of the two leading nations together meant that
three out of every four slaves carried to the Americas sailed in either a British
or a Portuguese vessel (including colonial-based vessels), and if France is
added, the ratio of the top three rises to nine out of ten. The half dozen other
nations in the trade, including the Netherlands, were minor carriers. The
major slave trading powers were, unsurprisingly, also the three nations domi-
nating the Atlantic commodity trades, though of the three, Portugal had much
the greater share of its total Atlantic trade accounted for by the slave trade and,
indeed, a much larger share of its total income generated by slave-related activ-
ities than the other two. Most minor traders, including the Dutch, came late
to the trade, managed to establish a position when the business was expand-
ing rapidly through to its peak in the second half of the eighteenth century,
and were the first to leave when the rate of expansion leveled off or when
absolute decline set in after the 1780s. The minor flags that appear in the
slave trade after i820 were in most cases flags of convenience, with the real
owners being mainly Spanish and Portuguese.

With the overall dimensions of the traffic established, the next step is to
distribute the estimated national totals of slaves by African region.53 Eight
regions of Africa are identified in the accompanying map (see Map II). An
important preliminary question is how often vessels obtained slaves at only
one point on the coast. The short answer is that trading was limited to one
region much more often than is commonly thought. When sending ships to
the African coast, merchants selected cargoes for specific markets because
Africans had regionally distinct preferences for merchandise. Ninety-five
percent of the cowries carried to the coast went to Bight of Benin ports.
Almost all the metal shipped from Europe to Africa went to Senegambia or
the Bight of Biafra, and manillas (wristlets) would sell only in the latter
region. Almost all New England rum sold on the Windward, Sierra Leone,
and Gold Coasts and all roll tobacco from Bahia went to the Slave Coast.
Textiles were welcomed in many places, although a pattern and texture that
would sell in one place often would sell nowhere else. Ships leaving on a
slave voyage would normally trade in only one region, though sometimes at
several locations in that region. Only 11.5 percent of vessels in the dataset
traded at two or more places, and 5.2 percent (8i2 voyages) of all records
with information on place of trade traded across regional boundaries. There
is undoubtedly some downward bias in these figures, in that vessels heading

53 The technical problems are straightforward in spreadsheet terms. Nevertheless, the pro-
cedures do involve considerable detail and some inferences about which means are appropriate.
It is possible for researchers to generate slightly different distributions from those presented
below on the basis of different (and equally plausible) assumptions.

This content downloaded from


139.47.113.28 on Fri, 07 Jan 2022 09:38:16 UTC
All use subject to https://about.jstor.org/terms
32 WILLIAM AND MARY QUARTERLY

for major West African regions east of the Cape Palmas typically acquired a
few slaves as they touched at points on the Windward Coast en route, with-
out reporting the fact later in the voyage. In addition, in some regions,
coastal lagoons or rivers made possible the movement of slaves along the coast
before embarkation. But even in the Bight of Benin, which had the most
extensive internal waterways, the movement of slaves from one region (as
opposed to port) to another was not cheap and cannot have been extensive.
On the African coast, trading for slaves was a highly specialized busi-
ness. Particular European captains would return to the same region many
times to trade with the same merchants. Particular European ports would
trade with the same African ports for many years and switch to new areas
reluctantly. Less often, particular European nations would dominate the
trade with entire African regions. Thus, all nine of Stephen Deane's slaving
voyages from London were to the River Gambia, and on the ninth he carried
not only slaves but also a female African-born merchant who emigrated as a
free person to Charleston and with whom he was clearly on good terms.54
Patrick Fairweather's fifteen voyages from Liverpool were all to Old Calabar
and the Cameroons, and he was one of many Liverpool captains who traded
for slaves only at those locations.55 Arend Bruyn sailed nine times from
Amsterdam to the Gold Coast between 1746 and 1763. Manoel Joaquim
d'Almeida captained eleven voyages from Bahia to the Bight of Benin, prob-
ably all to Lagos, between i814 and i826.56 Four out of five Rhode Island
vessels obtained their slaves from a short stretch of the Gold Coast. In the
Bight of Biafra, London vessels accounted for three out of four slaves carried
to the Americas between i650 and I700; Bristol, 70 percent of the slaves
taken between 1726 and I750; and Liverpool, 73 percent of those leaving in
the last quarter of the eighteenth century. In the 1780s, French vessels car-
ried 85 percent of the slaves from Congo River outlets. In the same decade,
all vessels leaving the southern outlets of West Central Africa (Luanda and
Benguela) were Portuguese, increasingly, from Rio de Janeiro. A slave vessel
rarely ventured to more than one African region to obtain its slaves.
The dataset contains large samples of African points of slave purchase
or, in the case of vessels that did not make it to the coast, the intended desti-
nations for each group of national traders. The African point of embarkation
is known for 62.6 percent of the English vessels, 76.1 percent of the French,
76.5 percent of the Portuguese, 24.8 percent of the Spanish, 45.7 percent of
the Dutch, 43.5 percent of United States voyages, and 52 percent of those
(mainly Danish) sailing from northern Europe. Overall, some indication of
African place of trade exists for nearly two out of three slave voyages or,
allowing for missing voyages, perhaps half of all transatlantic ventures that

54 Lillian Ashcraft-Eason, "She 'Voluntarily Hath Come to ... [the Georgia] Province': A
Gambian-Woman Slave Trader Among the Enslaved," in Lovejoy, ed., Constructions of Identity:
African Communities in the Shadow of Slavery (forthcoming).
55 See Behrendt, "The Diary of Antera Duke and the Eighteenth-Century Old Calabar
Slave Trade to the Americas," paper presented to the Michigan Seminar for Colonial Studies,
Oct. i6, i998.
56 Eltis et al., eds., Trans-Atlantic Slave Trade Database; Verger, Trade Relations, 404.

This content downloaded from


139.47.113.28 on Fri, 07 Jan 2022 09:38:16 UTC
All use subject to https://about.jstor.org/terms
VOLUME AND STRUCTURE OF THE SLAVE TRADE 33

obtained slaves in Africa. Such a large sample permits reliable estimates of


African coastal involvement, particularly because, except for the very early
period, it is evenly distributed over time. Moreover, the large gaps that exist
in our knowledge are usually in particular regions, and if there is solid infor-
mation available for other regions, sensible inferences are possible. Thus,
whereas little is known of the Portuguese traffic in the second half of the
seventeenth century, non-Portuguese sources for all African regions except
West Central Africa are fairly abundant, at least after i675. Such sources
show no activity on the part of Rio de Janeiro-bound Portuguese vessels.
What Rio trade existed must have been concentrated in Angolan ports.
Notable patterns in the departure of slaves from African regions are
apparent in Table II. The slave trade did not evolve gradually, or indeed
decline, in all areas of Africa at once. The sheer human and environmental
diversity of sub-Saharan Africa made this outcome unlikely. Although the
overall dominance of West Central Africa in the transatlantic slave trade is
well recognized, it is not generally appreciated that for the first 150 years the
region supplied nine out of ten slaves crossing the Atlantic. Except for the
two quarter centuries, i676 to 1725, when the appropriately named "Slave
Coast," or Bight of Benin, was the leading region of departure, West Central
Africa sent more slaves to the Americas than any other region in every quarter
between 15ig and i867. During 350 years, the center of gravity of the Atlantic
slave trade moved slowly north, away from West Central Africa, and then in
the mid-eighteenth century began an equally pronounced reverse movement.
For a century at the beginning and a quarter century at the end of the trade,
West Central Africa dispatched more slaves than all other regions combined,
and between i8oo and I850 it came close to doing the same.
Every African region experienced a single marked rise in numbers of
people pulled into the trade, followed by a plateau.57 West Central Africa
was the first to experience this large upswing probably at the end of the six-
teenth century when the slave trade to Brazil and the Spanish Americas
became substantial. Slavery spread slowly in the Americas initially, and it
was not until a century later that expansion of sugar output, mainly in the
English Caribbean, pulled the Bight of Benin into the slave trade orbit.
Further expansion, this time in both the Caribbean and Brazil, drew the
Gold Coast into the maelstrom after I700, followed by the Bight of Biafra
after I725, the three Upper Guinea regions after I750, and lastly, as the
Brazilian coffee and sugar boom got under way, Southeast Africa at the end
of the eighteenth century. After the initial upsurge, most regions experienced
high levels of departures for a century or more. Upper Guinea is an excep-
tion to this pattern. The Windward Coast remained at these higher levels for
just two decades, Senegambia for somewhat longer, and Sierra Leone for half

57 West Central Africa appears as an exception to this rule, but in reality it was very much
part of the general pattern. First, the pre-i65o intervals in Table II are for 50 years or more,
whereas those post-i65o are for 25 years only. Second, the present data are for transatlantic trad-
ing only. Slave vessels carried slaves to Old World ports in earlier years, and, though the num-
bers were far below what they were to become in the traffic to the Americas, a disproportionate
share of these departures were from West Central Africa.

This content downloaded from


139.47.113.28 on Fri, 07 Jan 2022 09:38:16 UTC
All use subject to https://about.jstor.org/terms
34 WILLIAM AND MARY QUARTERLY

a century. Except for the third quarter of the eighteenth century, these three
regions together were fortunately never able to match the volume of slaves
leaving any one of the other five regions and for most periods should be
considered minor suppliers of slaves to the Americas. Given the relative
proximity of Senegambia and the Windward Coast to the Americas-pas-
sage from Senegambia to the Caribbean was typically half as long as from
further south-this marginality is remarkable and suggests factors other than
geography at work.58
Both the sequence of regional participation and the pronounced and
widespread stepped pattern of expansion raise interesting questions about
the means and capacity of regions to supply slaves. It is likely that, apart
from West Central Africa, which sent ever larger numbers into the trade
after its initial expansion, the slave trade advanced in Africa mainly by push-
ing regions across a threshold of supply. A large surge of departures marked
the crossing of this threshold followed by a slightly lower plateau of depar-
tures. The intriguing issue is the cause and meaning of the step-up in slave
supplies. In each region, it occurred after many years when slave departures
averaged only a few hundred a year. One possible explanation is the time,
resources, and adjustment of social structures required to establish a supply
network or to break through to new sources in the interior or perhaps to
redirect an existing network for goods or slaves toward the Atlantic.
The particular events marking the step-up were different in each region.
The large increase in the number of slaves leaving the Bight of Biafra, for
example, coincides with the rise to influence of the Aro in the Niger Delta
and Cross River hinterlands.59 On the Gold Coast, gold was initially the
preferred commodity for both Africans and Europeans, and the massive
increase in slaving in this region coincides with the exhaustion of deposits
and the reorientation of trade routes from gold to slaves.60 In the Bight of
Benin, the key event appears to have been the reorganization of the trade
associated with the emergence of Whydah. From the perspective of long-run
trends in the number of slaves sent to the Americas, the rise of Whydah was
far more significant than the subsequent conquest of Whydah by the much
larger state of Dahomey. Incorporation into a larger polity was in this case
associated with a one-third drop in slave departures in the twenty-five years
after conquest. Only in West Central Africa did the numbers continue to
grow dramatically after the initial massive increase; unlike in other regions, a
continually shifting frontier of slaving apparently existed there.61 In West
Central Africa, the outcome of increasing demand and rising prices was an
expansion inland of sources of slaves-no doubt a function of the huge size

58 Of the 3 Upper Guinea regions, Sierra Leone sustained the higher departure levels for
longer and came closest to following the pattern of regions to the south. For a fuller discussion
of this pattern see Behrendt, Eltis, and Richardson, "The Costs of Coercion: African Agency in
the Atlantic World," Economic History Review (forthcoming).
59 G. Ugo Nwokeji, "The Biafran Frontier: Trade, Slaves and Aro Society" (Ph. D. diss.,
University of Toronto, i999), esp. 38-66.
60 For a discussion and review of the literature see Eltis, Rise of African Slavery, I50-5I.
61 Miller, Way ofDeath, I40-53, 234-4I.

This content downloaded from


139.47.113.28 on Fri, 07 Jan 2022 09:38:16 UTC
All use subject to https://about.jstor.org/terms
VOLUME AND STRUCTURE OF THE SLAVE TRADE 35

of the hinterland of what were essentially three outlets to the Atlantic-the


Congo, Luanda, and Benguela. In all other areas of Africa, the trade
expanded by pulling in other regions of the coast after the first dramatic
increase in a single region.
Interestingly, the decline of the slave trade was a mirror image of its
expansion-a series of regional steps. Numbers declined first on the
Windward Coast after I780, then in Senegambia and the Gold Coast after
abolition by the United States and Britain. The Bight of Biafra followed at
the end of the I840s, and Sierra Leone thereafter. The transatlantic part of
southeast African departures fell precipitously after I85I, which left only th
Bight of Benin and West Central Africa to supply Cuba until i867.62

In the Americas, more ports and a wider geographic range of regions


were associated with the slave trade than in Africa. The destination or slave
disembarkation point is known for 84.4 percent of English voyages, 8o.i per-
cent of the French, 62.3 percent of the Portuguese, 90.6 percent of the
Spanish, 72.3 percent of the Dutch, 63.7 of United States voyages, and 75.8
percent of those (mainly Danish) sailing from northern Europe. Overall,
some indications of region of disembarkation exists for more than three out
of four slave voyages in the dataset, and as with the distribution across
African regions, it is possible to apportion totals of slaves carried by each
national group across regions of disembarkation. Sixteen regions of disem-
barkation are used, fifteen in the Americas-mainland British North
America (including Mississippi), the British Windwards and Trinidad
together, the British Leewards, Jamaica, Barbados, all the Guianas together,
mainland Spanish America, the Spanish Caribbean, Brazil north of Bahia,
Bahia, Brazil south of Bahia, the Dutch Caribbean, the French Windwards,
St. Domingue, and finally, all other Americas, 90 percent of which com-
prises the Danish islands. For reasons of distress or capture (particularly the
latter when efforts to suppress the trade began), many vessels disembarked in
Africa, and the continent that supplied the slaves becomes also the sixteenth
region of disembarkation. Table III provides the results.
Perhaps 5-Io percent of all slaves who arrived in the Americas quickly
moved to other parts of the Americas, as part of an intra-American slave
trade. Indeed, for nearly I5o years after i65o, slaves arrived in most of Spanish
America via entrepots rather than directly from Africa. Unfortunately, such
movements are not yet captured in the dataset, which is based on the transat-
lantic slave trade only. As a proportion of the total Atlantic traffic in human
beings, the intra-American trade does not seem large, nor was it evenly dis-
tributed. Regions such as Guadeloupe and eighteenth-century Mississippi
obtained most of their slaves from other parts of the Americas rather than
direct from Africa. The mainland Spanish colonies may have received half of
their arrivals, and the mainland British colonies fewer than 5 percent in this
way. The suppression of the transatlantic slave trade triggered a much greater
intra-American traffic (and probably intra-African traffic, too), as slaves were

62 Eltis, Economic Growth, i64-84.

This content downloaded from


139.47.113.28 on Fri, 07 Jan 2022 09:38:16 UTC
All use subject to https://about.jstor.org/terms
36 WILLIAM AND MARY QUARTERLY

carried from the Old South to the New in the United States, from northeast
to southeast Brazil, and on a much more restricted basis from the old British
Caribbean to newly acquired British territories.63 Some attempt is made to
adjust for the preabolition patterns below, but less is known about this early
phase of the intra-American trade.
The major adjustments to Table III the intra-American trade make nec-
essary are to mainland Spanish America-the major recipient of slaves trans-
ported in the intra-American traffic. Perhaps 85 percent of the slaves arrivin
in the Dutch Caribbean were quickly re-embarked for ports on the Spanish
Main. Jamaica was an even larger market or transit point for Spanish buyers,
whereas Barbados played a supplementary role in the seventeenth century,
and an estimate of shares passed on to the Spanish is more difficult for these
markets. Spanish demand for slaves from the mainland focused on Jamaica
after I700, falling off late in the eighteenth century, and except for
I73I-I735, is unlikely to have accounted for more than 30 percent of Afric
arriving in that island for any five-year period.64 Perhaps on average, IS per-
cent of pre-I700 Jamaican arrivals (rising to 20 percent from I700 to I808),
and 4 percent of pre-I7I5 arrivals in Barbados were shipped to Cartagena,
Vera Cruz, or Portobello and the Caracas coast. In addition there was an
intermittent trade from southeast Brazil to the Rio de la Plata down to the
i83os that introduced far fewer slaves than the direct traffic from Africa.65 In
Table III, mainland Spanish America accounts for 4.4 percent of the slave
trade. Adjustment for the intra-American trade might raise this share to 7 or
8 percent, mainly at the expense of the Dutch Caribbean and Jamaican
columns in the table. Mainland British North American colonies received
fewer intra-American arrivals and these in much smaller groups than did the
Spanish. In the Chesapeake, they accounted for 7 percent of all slaves enter-
ing and in South Carolina just 3.8 percent, though these figures run counter
to the perceptions of some historians.66 If as many as one-tenth of slave

63 The best estimates are in Robert W. Fogel and Engerman, Time on the Cross: The
Economics of American Negro Slavery, 2 vols. (Boston, I974), I:44-52; Robert W. Slenes, "The
Demography and Economics of Brazilian Slavery: I85o-I888" (Ph. D. diss., Stanford University
I975), I20-79, esp I27, I37; and B. W. Higman. Slave Populations of the British Caribbean
(Baltimore, I984), 79-88, 429-30. For the intra-African trade in southeast Nigeria in the after-
math of suppression of the Atlantic traffic, see Nwokeji, "Biafran Frontier," 200-II.
64 The best series for Jamaican sales of slaves out of that island, together with a discussion
of alternative sources, is Richard B. Sheridan, "Slave Demography in the British West Indies
and the Abolition of the Slave Trade," in Eltis and James Walvin, eds., The Abolition of the
Transatlantic Slave Trade: Origins and Effects in Europe, Africa, and the Americas (Madison,
I981), 274-75. Sheridan warns of smuggling, but his figure of I93,597 is I9.7% of the I70I-I808
estimate of arrivals in Jamaica in Table III of the present article.
65 Elena F. S. de Studer, La trata de Negros en Rio de la Plata durante el siglo XVIII (Buenos
Aires, I958), insert, quadro 5.
66 Lorena S. Walsh demonstrates that 93% of Chesapeake arrivals came directly from
Africa, in "The Chesapeake Slave Trade: Regional Patterns, African Origins, and Some
Implications," in this volume. The South Carolina ratio is calculated from Eltis et al., eds.,
Trans-Atlantic Slave Trade Database, for arrivals from Africa and from Elizabeth Donnan, ed.,
Documents Illustrative of the History of the Slave Trade to America, 4 vols. (Washington, D. C.,
I930-I933), 4:296-58I, for arrivals from the Caribbean. The ratio is for 34 select years, I739-I807.

This content downloaded from


139.47.113.28 on Fri, 07 Jan 2022 09:38:16 UTC
All use subject to https://about.jstor.org/terms
VOLUME AND STRUCTURE OF THE SLAVE TRADE 37

arrivals into what became the United States originated in the Caribbean,
then total slave arrivals into mainland North America would have been less
than 400,000. This is probably the lowest estimate of numbers of slaves
arriving in mainland North America ever made. Other small adjustments
required for the intra-American trade are harder to quantify. Except when
the English temporarily held the island toward the end of the Seven Years'
War, Guadeloupe received almost all its slaves via Martinique, though Table
III groups the islands in the same column. The French Windwards as a
whole were supplied from Dutch and English islands as well as from Africa,
and the trade probably averaged a few hundred per annum before I720. In
addition, both Barbados and St. Eustatius supplied part of the English
Leewards inflow down to I730, and Grenada became a significant entrep t
late in the century, but, again, these were not large trades. Finally, the move-
ment of English slaves into Demerara began before the British took over the
colony from the Dutch in I796.
None of these adjustments, including those for the Spanish and British
mainland colonies, substantially reconfigures the overall picture of arrivals
into the American regions shown in Table III. Given the dominance of the
Portuguese and British slave trades, it is hardly surprising that Brazil and
British America garnered most Africans. The former accounts for 4I percent
of the slaves arriving in the New World and the latter for 29 percent.67 After
making best guesses at the subsequent movement of slaves to other jurisdic-
tions, about seven out of every ten slaves carried across the Atlantic remained
in Brazil or the English-speaking Americas (though with the mainland claim-
ing a larger share than shown in Table III). The French Americas absorbed
half the number of slaves that the British Americas did, with St. Domingue
taking in fewer slaves than Jamaica, after adjusting for Spanish buyers, even
though the largest French island (assumed to be St. Domingue) far outpro-
duced its English counterpart. The French Americas actually took in fewer
slaves than any of the major European empires in the Americas. The explana-
tion is that St. Domingue, which accounted for nearly 8o percent of all
slaves coming to the French Americas, was a major plantation colony for a
relatively short time. Insignificant at the beginning of the eighteenth cen-
tury, in ruins at the end, and the colony of a country usually on the losing
end of wars that disrupted the traffic in slaves, it nevertheless took in more
slaves than Jamaica between I75o and I79I. The British Americas were more
extensive geographically and began to take large numbers of slaves earlier
than the French islands. The trade to the Spanish mainland began earlier
than to anywhere else and, through Cuba, grew explosively even while the
trade to the rest of the Americas was shutting down.
Assessments of the relative importance of regions in the Americas to the
slave trade have remained substantially unchanged since Curtin's census, and

Many arrivals from the Caribbean are identified in these tables by African coastal region, suggest-
ing that the slaves had not been in the Caribbean for very long and that the ratio may be biased
upward.
67 After breaking up the Guianas total among French, Dutch, and British.

This content downloaded from


139.47.113.28 on Fri, 07 Jan 2022 09:38:16 UTC
All use subject to https://about.jstor.org/terms
38 WILLIAM AND MARY QUARTERLY

Table III is the first comprehensive modification to his thirty-year-old pic-


ture. Brazil and the British Caribbean come out as more important and the
French Caribbean and Spanish Americas less important than before, whether
we use the ratios in the last row of the table or the crudely adjusted figures,
which attempt to take the intra-American trade into account. Brazil now
appears to have taken 40.6 percent of all Africans, instead of the 39.0 per-
cent that Curtin estimated. The British Americas absorbed 29.0 percent (just
over 27.0 percent after adjusting for sales to the Spanish), up substantially
from Curtin's unadjusted 2I.6 percent. The small trade to the Danish islands
is now reckoned to have been about i percent, up from Curtin's 0.3 percent.
These upward revisions are at the expense of the Spanish Americas, which
took in an adjusted I4.3 percent instead of Curtin's i6.3 percent, the French
colonies (except Cayenne) I2.0 rather than I5.7 percent, and the Dutch
Americas, 2.7 percent instead of 5.2, excluding those put into the intra-
American traffic. Over time, more slaves came in the nineteenth century and
fewer before i66o than Curtin posited. Nevertheless, the differences between
Curtin's Census and Table III cannot be considered large.
The new distribution does underscore the importance of sugar in the
shaping of the transatlantic slave trade. Slaves crossed the Atlantic for forced
labor primarily in sugar production or the economies that sugar exports
made possible. The tobacco farmers of the Chesapeake and Bahia created a
small market in the broad scheme of the Americas, as did the rice growers of
Carolina. Coffee accounted for just over one-third the value of sugar exports
from the plantation Americas in I770. Cotton became a major crop only as
the transatlantic traffic to the United States was ending. Indigo and cacao
were minor sectors overall though locally important. Gold and silver produc-
tion in the early Spanish Americas and Minas Gerais in Brazil after i695 was
an important source of demand. Even though Spanish America dominated
the early slave trade, the flow of slaves to this area was small compared to
what it became, and, although gold discoveries in Minas Gerais are credited
with the large expansion of the Brazilian trade in the early eighteenth cen-
tury, far more slaves-both in total and per year-arrived after the gold
boom subsided. Moreover, in both Spanish America and Brazil, specie booms
were associated with the arrival of more people from Europe than from
Africa, whereas in the sugar colonies in the two centuries after i65o increased
sugar exports almost invariably meant far greater African arrivals than
European. In both Brazil and Spanish America, most slaves ended up work-
ing in a variety of employments made possible by the new prosperity, rather
than directly in the mines or placer operations that produced the precious
metals. Nevertheless, competition for sugar planters did not emerge not until
late in the slave trade era. Coffee in southeast Brazil, more specifically the
Paraiba Valley, created the largest single market for transatlantic slaves in the
Americas after I820. But from the late sixteenth century down to I820, 90
percent of the slaves were brought across the Atlantic because of sugar.68

68 Fogel, Withont Consent or Contract: The Rise and Fall of American Slavery (New York,
I989), I8-2I.

This content downloaded from


139.47.113.28 on Fri, 07 Jan 2022 09:38:16 UTC
All use subject to https://about.jstor.org/terms
VOLUME AND STRUCTURE OF THE SLAVE TRADE 39

The most exciting implications of the new dataset are not the volume
and distribution of the trade on either side of the Atlantic, but rather the
new light it casts on transatlantic connections. For the first time, estimates of
the African origins of slaves arriving in almost any part of the Americas, as
well as of the American destinations of slaves leaving the major regions of
Africa, are available from large samples of shipping data. These linkages
show that the distribution of Africans in the New World was no more ran-
dom than the distribution of Europeans. A couple of major admitting areas
received slaves from mainly one part of Africa for most of the trade. Four out
of five slaves taken to southeast Brazil came from West Central Africa, and
further north, Bahia drew a similar proportion from the Bight of Benin after
i68o, Adja-speaking peoples before I780, Yoruba thereafter.69 At the other
extreme Cuba received the greatest mix of African peoples at the same time.
No single part of Africa supplied more than 28 percent of arrivals to Cuba,
and the only major region not well represented was the Gold Coast. In addi-
tion, the region that supplied the greatest number of slaves-West Central
Africa-drew on a longer coastline and a larger hinterland by the nineteenth
century. Nor was there any regional segregation in Cuba. Four-fifths of
arrivals moved through Havana and other western ports and for the rest of
their lives worked in the sugar heartland of the island. Large numbers of Yao
from the southeast, Yoruba from West Africa, and Lunda from the Kasai
Valley in the Angolan interior-almost all of whom arrived in the first half
of the nineteenth century-worked in the plantation labor forces.
The pattern of transatlantic links for most of the Americas falls between
these two extremes. Typically, an American region drew on a mix of regions
yet tended to do so in sequence, a sequence moreover that in the British
islands played out slowly. Over time there was also a tendency to draw on a
more diverse set of provenance zones simultaneously. In sum, in the early
period of slavery in the Americas, ethnicity for persons of African descent
was rather more exclusive than it became in the last quarter of the eighteenth
century. Thus, in the Guianas, two-thirds of all arrivals before I750 came
from the adjacent Gold Coasts and Bight of Benin. A switch to West Central
Africa in the third quarter was followed by a Cuban-style provenance pattern
after I775. In the Rio de la Plata, nearly 90 percent of Africans arriving
before I780 came from the southern ports of West Central Africa, the mix
thereafter excluded only Upper Guinea. The Danish islands had an almost
exclusive link with the Gold Coast and the Bight of Benin until I775; then,
West Central Africa and the Bight of Biafra supplied most of their slaves.
Interestingly, Danish vessels carried amost all the slaves before I775, whereas
vessels belonging to other nations predominated in the period of more
diverse provenance. The same regions supplied Barbados between i670 and
I725, after which the Bight of Biafra replaced the Bight of Benin for fifty
years, and Angola replaced the Gold Coast after I775. In the French
Windwards, three out of five slaves arrived from a few ports in the Bight of

69 Manning, Slavery Colonialism and Economic Growth in Dahomey, 1640-1960 (New York,
I982), 29-32.

This content downloaded from


139.47.113.28 on Fri, 07 Jan 2022 09:38:16 UTC
All use subject to https://about.jstor.org/terms
40 WILLIAM AND MARY QUARTERLY

Benin before I750, when plantations in th


ing the region most rapidly. Thereafter,
Biafra (I776-I792) supplied nearly two out of five. The British Leewards
drew well over half their slaves from the Bight of Biafra before I775, bu
during the rest of the slave trade era there was no single dominant region.
No Upper Guinea region had a major role in supplying any American
region, the good anchorages and infrastructure of Goree, Fort James in the
Gambia, Bance Island, and Sierra Leone notwithstanding. In most regions,
the long-run effect was a mixing of peoples, but, particularly before I750,
the less diverse origins of coerced early African migrants held.
There are intriguing parallels with European settlement. With the
exception of the Spanish, most European settlements in the Americas began
with migrants from a single area and, especially in the English, Dutch, and
Portuguese slave colonies, quickly came to have a much greater range of
European occupants than their metropolitan owners. Rights of denizenship,
even for Jews, were easier to obtain in the colonies than in the home coun-
try-Spanish Americas excepted. Perhaps the African migrant populations in
the Americas were no more mixed than peoples of European descent in the
Americas. Formal transatlantic colonial and commercial ties have obscured
the polyglot nature of Euro-American societies and at the same time discour-
aged the search for geographic links between Africa and the Americas.
Fuller analyses of two major colonies, Jamaica and St. Domingue, will
illustrate the relative African origins of slaves (see Tables IV and V).
Jamaica's intra-American trade as well as lack of slaves before i672 must be
taken into account in interpreting Table IV. As the factors of the Royal
African Company frequently pointed out, in the i672-I7I3 period a highly
disproportionate share of slaves from Senegambia, the Bight of Biafra, and
West Central Africa arriving in Jamaica were sold to Spanish buyers and left
the island.70 Given this pattern of resale and given that the predominant
Bight of Biafra provenance of the i66os (row i, Table IV) occurred when
Jamaica arrivals were only a thousand a year before resale, three quarters of
all slaves retained in Jamaica before I725 probably were from the adjacent
Gold Coast and Slave Coast. The pattern of arrivals over the next eighty
years is one of a dramatic fall in the importance of the Bight of Benin, a
gradual decline in Gold Coast arrivals, and the rise to importance of Bight
of Biafra and, to a lesser extent, Angola-essentially, the Congo. By the end
of the slave trade era, nearly two out of three slaves-most of them Igbo
from the Bight of Biafra-were coming from these last two areas, but nei-
ther region became as important as the Gold Coast had been.
In St. Domingue, three out of five arrivals, mostly Adja speakers, came
from the Bight of Benin alone before I725, with Senegambia accounting for
most of the rest (see Table V). Both regions declined in importance rather
quickly thereafter, so that far more than half of all slaves came from Angola
after I725. The variety of peoples among new arrivals from other regions was

70 Eltis, Rise ofAfrican Slavegy, chap. 9.

This content downloaded from


139.47.113.28 on Fri, 07 Jan 2022 09:38:16 UTC
All use subject to https://about.jstor.org/terms
VOLUME AND STRUCTURE OF THE SLAVE TRADE 4I

much greater after I750 than before, and with the switch in the dominant
African region, the ethnicity of the slave population by the end of the period
was clearly more mixed. Historians should not expect to find pronounced
evidence of Igbo, Akan, and Yao in St. Domingue, in strong contrast to
Jamaica. Because the island's links with the Bight of Benin predated the fall
of the Oyo empire (with its disastrous results for the Yoruba), Adja influence
should have been (and indeed was) much more in evidence than Yoruba. The
broader implications of these patterns for African-American life is beyond
the scope of this article, but one finding is that the ethnic environment of
early Jamaica and St. Domingue was much less diverse than it was to become
later. A second observation is that the patterns observed here do not accord
with planters' expressed preferences. There were clearly other factors driving
the slave trade.71

A number of provisional conclusions on the impact of the new dataset are


possible for the general structure of the trade. First, Philip Curtin was close to
the mark in his estimate of the overall volume of the slave trade. His Atlantic
Slave Trade. A Census is a rare academic publication that has aged well. On the
participation of European nationals, two nations dominated. Before i650, the
Portuguese transported more than 95 percent of what appear by later standards
to be a small flow of slaves. Between i66o and I807, when the slave trade was at
its height, the British and their dependencies carried every second slave that
arrived in the Americas, a dominance that would no doubt have continued but
for the politically inspired decision to abolish the trade. After the British
dropped out, the Portuguese resumed their previous position though not to
quite the same extent. The French needed subsidies to compete; the Dutch, the
Danes, and (except for the last two decades before abolition) the Americans had
minor roles, and the Spanish became important only when almost everyone else
pulled out of the traffic. On the African coast, West Central Africa was an even
more important source for slaves than the recent literature credits; a swing away
from, then a return to Angola occurred. More important, for every region out-
side Angola a marked rise in slave departures occurred in sequence, followed by
a plateau of departures that continued until a rather sudden end to the traffic.
In the Americas, sugar was the driving force in the slave trade, though gold and
silver were important in the earliest phase of the traffic, and coffee assumed that
role in the final phase. Except for the meteoric rise and fall of St. Domingue,
the primary receiving regions were Brazil, the British Caribbean, and, briefly in
the nineteenth century, Cuba. Finally, on the issue of transatlantic links, the
picture of African coerced migrants arriving mainly in a mix of peoples-often
on the same vessel-needs revising. Like the free migrant and indentured ser-
vant trades, systematic geographic patterns existed. Scholars should now turn to
exploring what these mean both for Africa and for African influences in the
shaping of the New World
The direction of this exploration is suggested by the experience of
assembling and working with the new dataset and responding to the numer-

7I See Walsh, "Chesapeake Slave Trade."

This content downloaded from


139.47.113.28 on Fri, 07 Jan 2022 09:38:16 UTC
All use subject to https://about.jstor.org/terms
42 WILLIAM AND MARY QUARTERLY

ous scholars who analyzed sections of it before it was complete. Time and
again, as the articles in this journal suggest, the data reveal patterns that are
not easy to explain if the focus is on Europe and the Americas or on Europe
and the Americas alone. Why, for example, was the area closest to Europe
and the Americas, with the best natural harbors and by far the shortest
Middle Passage (Upper Guinea), the area that sent fewest slaves to the
Americas? Whether the topic is European merchant decisions to launch a
voyage as well as when and where to send it, African resistance, regional
mortality rate differentials, age and sex profiles, the volume of the trade, or
links between African and American regions, understanding is enhanced if
Africans are treated as more than mere victims. Thus, if the current findings
have any general implications, they are, first, that a larger role for African
agency in interpretations of the shaping of the trade is essential, and, second,
that an Atlantic framework that allows for interaction among four continents
is required. These two points in no way contradict each other. In modern
terms, the slave trade was a massive and enduring crime against humanity:
slave ships never carried white slaves, and very few transatlantic slave voyages
were organized in Africa, but if these were to be the only conclusions carried
away from the new dataset, then it will have failed to advance our under-
standing of how the slave trade could have happened and what it means in
human history.

This content downloaded from


139.47.113.28 on Fri, 07 Jan 2022 09:38:16 UTC
All use subject to https://about.jstor.org/terms
TABLE I
VOLUME OF THE TRANSATLANTIC SLAVE TRADE FROM AFRICA BY NATIO

Britain France Netherlands Spain United States De


and British
Caribbean

1519-1600 2.0 264


1601-1650 23.0 41.0 4
1651-1675 115.2 5.9 64.8 0.
1676-1700 243.3 34.1 56.1 15.
1701-1725 380.9 106.3 65.5 11.0 1
1726-1750 490.5 253.9 109.2 44.5 7
1751-1775 859.1 321.5 148.0 1.0 89.1 1
1776-1800 741.3 419.5 40.8 8.6 54.3 3
1801-1825 257.0 217.9** 2.3 204.8** 81.1 1
1826-1850 94.1 ** 279.2** 1,24
1851-1867 3.2** 23.4** 15
All Years 3,112.3 1,456.4 527.7 517.0 280.0 9
Share ofTrade 28.1% 13.2% 4.8% 4.7% 2.5%

Source: David Eltis, Stephen D. Behrendt, David Rich


(Cambridge, i999), adjusted according to text.
* Denmark includes vessels from a few other Scandinav
**For the i8i0-i867 period, shares of slave vessels sailing
estimates of slave departures and arrivals (2.7 million depa
Trade (New York, I987), 250-54. The amounts are: Frenc
i,944,600 departures, i,692,5oo arrivals; and Spanish ships,
French engages taken during the I85os and i86os are not i
Totals have been rounded; for exact numbers, see <http

This content downloaded from


139.47.113.28 on Fri, 07 Jan 2022 09:38:16 UTC
All use subject to https://about.jstor.org/terms
TABLE II
VOLUME OF TRANSATLANTIC SLAVE DEPARTURES BY REGION

Senegambia Sierra Leone Windward Coast Gold Coast Bight of Benin Bight o

1519-1600 10.7 2.0 10.7 10.7 10


1601-1650 6.4 5.2 2.4 25.5 46
1651-1675 17.7 0.4 0.1 35.4 21.9 58.
1676-1700 36.5 3.5 0.7 50.3 223.5 51.5
1701-1725 39.9 7.1 4.2 181.7 408.3 45.8
1726-1750 69.9 10.5 14.3 186.3 306.1 166.
1751-1775 130.4 96.9 105.1 263.9 250.5 340
1776-1800 72.4 106.0 19.5 240.7 264.6 360.
1801-1825 91.7 69.7 24.0 69.0 263.3 260.3
1826-1850 22.8 100.4 14.4 257.3 191.5 7
1851-1867 16.1 0.6 25.9 7.3 15
AllYears 498.5 412.7 183.0 1,043.2 2,034.6 1,517.9
Share of Trade 4.5% 3.7% 1.7% 9.4% 18.4% 13.

Source: Eltis et al., eds., Trans-Atlantic Slave Trade Database, a


ken into five-year intervals (before i675, intervals of a quarter cent
summed by quarter century for each African region. Five-year brea

This content downloaded from


139.47.113.28 on Fri, 07 Jan 2022 09:38:16 UTC
All use subject to https://about.jstor.org/terms
TABLE III
VOLUME OF TRANSATLANTIC SLAVE ARRIVALS BY REGION OF ARRIVAL (IN THOUSANDS).

British Mainland British British Windwards Jamaica Barbados Guianas French


North America Leewards and Trinidad Windwards

1519-1600
1601-1650 1.4 1.0 0.2 25.4 2.0
1651-1675 0.9 5.6 22.3 63.2 8.2
1676-1700 9.8 26.6 73.5 82.3 27.8 16.6
1701-1725 37.4 35.4 0.6 139.1 91.8 24.4 30.1
1726-1750 96.8 81.7 0.3 186.5 73.6 83.6 66.8
1751-1775 116.9 123.9 120 270.15 120.9 111.9 63.7
1776-1800 24.4 25.3 197.5 312.6 28.5 71.2 41.2
1801-1825 73.2 5.3 43.0 70.2 7.6 71.8 58.8
1826-1850 0.5 2.1 0.9 4.8 19.5
1851-1867 0.3 0.4
All Years 361.1 304.9 362.0 1,077.1 494.2 403.7 305.2
Share of Trade 3.8% 3.2% 3.8% 11.2% 5.1% 4.2% 3.2

PART Z

Spanish American Spanish Dutch Northeast Bahia Southeast Other Africa


Mainland Caribbean Caribbean Brazil Brazil Americas

1519-1600 151.6 35.0 15.0 20


1601-1650 187.7 2.0 86.3 60.0 30.0 3
1651-1675 38.8 15.6 15.6 15.6 0.6 1
1676-1700 7.0 26.0 30.2 75.9 30.2 11.0
1701-1725 30.0 2.1 30.5 24.3 199.6 122.0 14.2
1726-1750 12.7 1.6 10.2 51.4 104.6 213.9 8.3 1
1751-1775 5.0 13.0 15.3 126.9 94.4 210.4 13.8 1
1776-1800 10.2 56.9 6.9 210.8 112.5 247.2 44.1 0.4 1
1801-1825 17.5 268.7 214.8 182.0 408.7 14.8 22.4 1
1826-1850 8.7 297.0 80.0 146.5 736.4 10.5 91.3 1
1851-1867 152.6 0.9 1.9 3.6 1.3 16.8 1
All Years 430.3 791.9 129.7 876.1 1,008.0 2,017.9 118.7 130.8
Share of Trade 4.5% 8.2% 1.4% 9.1% 10.5% 21.0% 1.2% 1.4%

Source: Eltis et al., eds., Trans-Atlantic Slave Trade Database, and


first broken into five-year intervals (before i675, intervals of a quarte
bution are then summed by quarter century for each American reg

This content downloaded from


139.47.113.28 on Fri, 07 Jan 2022 09:38:16 UTC
All use subject to https://about.jstor.org/terms
TABLE IV
AFRICAN REGION OF DEPARTURE OF SLAVES ARRIVING IN JAMAICA, B

Senegambia Sierra Leone Windward Gold Coast Bight of Benin Bi


Coast Afric

1651-1675 9.1 19.3 60.


1676-1700 6.8 1.4 10.6 34.8 16.1
1701-1725 5.3 1.1 1.0 45.7 33.8 1
1726-1750 3.1 1.2 1.0 27.2 9.2
1751-1775 2.0 4.8 10.9 31.8 10.8
1776-1800 0.9 4.2 3.1 25.3 6.5 4
1801-1808 0.4 2.2 2.4 21.2 4.8
1651-1808 2.0 3.5 4.7 27.7 11.8 33.
Source: Calculated from Eltis et al., eds., Trans-A

TABLE V
AFRICAN REGION OF DEPARTURE OF

Senegambia Sierra Leone Windw


Coast Afric

1701-1725 19.2 0.2 0.4 59.3 1.8


1726-1750 13.2 0.4 3.0 13.0 37.8
1751-1775 4.4 3.7 0.9 1.5 25.3 3.8
1776-1800 4.4 4.7 0.8 3.7 19.1 8.0
1701-1800 6.8 3.4 1.2 4.3 26.3 5.1
Source: Calculated from Eltis et al., eds., Trans-At

This content downloaded from


13ffff:ffff:ffff:ffff:ffff on Thu, 01 Jan 1976 12:34:56 UTC
All use subject to https://about.jstor.org/terms

You might also like